No.

420 December 13, 2001

Limiting Government through

Direct Democracy
The Case of State Tax and Expenditure Limitations
by Michael J. New

Executive Summary

Defenders of individual freedom and limited
government have often favored representative
government over direct democracy. Since 1978,
however, activists have proposed and passed ini-
tiatives limiting taxing and spending by state
governments in the United States. State legisla-
tures have occasionally imposed tax and expen-
diture limitations (TELs) on themselves.
TELs passed by initiative are more restrictive
and contain fewer loopholes than those enacted
by state legislatures. Regression analysis of a com-
prehensive data set of state government spending
shows that TELs enacted by citizen initiatives
cause per capita public spending to decrease; TELs
enacted by state legislatures are associated with an
increase in government expenditures.
Some TELs are more effective at limiting gov-
ernment than others. TELs that limit govern-
ment spending to the inflation rate plus popula-
tion growth and mandate immediate rebates of
government surpluses are more effective at limit-
ing government outlays than are other TELs.

_____________________________________________________________________________________________________
Michael J. New, a Ph.D. candidate at Stanford University, was a data analyst and research assistant at the Cato
Institute in 2001.
Initiatives can be
an effective tool Introduction
for limiting Direct democracy is on the rise again in
government. the United States. Since 1978 the initiative
has been used more and more in the 24 states
that permit direct legislation by voters. 1 In
California, for example, 9 voter initiatives
made it to the ballot in the 1960s; 22 made it
in the 1970s, 45 in the 1980s, and 62 in the
1990s. In the 1996 general election, American
voters were confronted with more than 90
statewide initiatives, plus an estimated 200
local initiatives and referenda on environ-
mental and land-use issues alone.2 As infor-
mation technology develops, direct and
immediate rule by the people may become a
real possibility.
Proponents of individual liberty and limit-
ed government have traditionally been skepti-
cal about direct rule by a majority of voters.
The reasons for that skepticism can be found
in The Federalist Papers and in the Progressive
Era’s push for bigger government through
direct democracy. The revival of direct democ-
racy began, however, with a 1978 California
initiative to reduce property taxes, the famous
Proposition 13 put on the ballot by Howard
Jarvis and Paul Gann.3 Many subsequent suc-
cessful initiatives imposed tax and expendi-
ture limitations (TELs) on state governments.
This paper will show that initiatives can be an
effective tool for limiting government. Theory
notwithstanding, direct democracy seems to
be a good friend of liberty in some cases.
Legislatures or the People?
James Madison had little faith in pure
democracy, “a Society, consisting of a small
number of citizens, who assemble and
administer the Government in person.” He
believed that direct democracy meant a
tyranny of a majority over minorities, the rule
of passion over reason. Pure democracy is
thus “incompatible with personal security, or
the rights of property.” The direct rule of the
people, Madison concluded, ended in fac-
2
tion, discord, and a violent death of the polit-
ical association.
Madison thought the new American Republic
overcame the ills of pure democracy. In part, his
confidence came from the scope of the American
Republic: by including more people and interests
in the nation, the Constitution made it harder for
dangerous majorities to form. Madison praised
also the representative character of the new
Constitution. Delegating power to legislators
would “refine and enlarge” public opinion, there-
by avoiding majority tyranny and perhaps attain-
ing higher goals: “It may well happen that the
public voice pronounced by the representatives of
the people, will be more consonant to the public
good, than if pronounced by the people them-
selves convened for the purpose.”4
The Progressives turned to the initiative
because they rejected both Madison’s belief in
representative institutions and his desire for
limited government. They thought that, far
from refining and enlarging public opinion,
legislatures “enact[ed] laws for the special
advantage of a few and refuse[d] to enact laws
for the welfare of the many.”5 Progressives saw
direct legislation as a way to bypass corrupt
legislatures and to expand the ambit of gov-
ernment. Benjamin Parke DeWitt in 1915 said
Progressives share “the rapidly growing con-
viction that the functions of government are
too restricted and that they must be increased
and extended to relieve social and economic
distress.”6 Progressives expected the initiative
to weaken corrupt politicians, empower the
people, and thereby expand government to
reform society. The initiative was instrumental
for attaining their larger goal of expanding
government.
Government did expand enormously dur-
ing the 20th century, but economic crisis and
war, not the initiative, were the most impor-
tant factors driving its growth.7 With govern-
ment growth came more and higher taxes at
all levels of government. Harry Hopkins’ dic-
tum “tax and tax, spend and spend, elect and
elect” became the practical wisdom of
America’s political class.
In 1978 Proposition 13 in California tried to
change that. The initiative concerned mostly
property taxes, which had been rising 30 per-
cent annually in California in the 1970s. It lim-
ited the assessed value of property, increases in
assessed property value, and the tax rate for
property. The proposition also required a
supermajority in the legislature to raise proper-
ty tax rates. At the time, experts thought that
the initiative would cut government revenue by
$7 billion in 1978–79. Approximately 69 per-
cent of registered voters turned out in a June
primary for Proposition 13; 65 percent of them
voted in its favor.8
Like the Progressives, the originators of
Proposition 13 believed the California legis-
lature and state government in general had
ignored the desires of citizens. They may have
been correct. A contemporary public opinion
poll showed that all subgroups of
Californians supported Proposition 13, with
the exception of blacks and “strong liberals.”
The legislature had not responded to the
broad-based desire for lower property taxes,
though it did pass a property tax cut after
Proposition 13 reached the ballot.9 In retro-
spect, Proposition 13 reflected the will of a
broad majority of voters in California in
1978, a desire ignored by the state legislators.
As tax cutters warmed to the idea of direct
rule of the people, the ideological heirs to
progressivism criticized the initiative.
Washington Post columnist David Broder
warned about the danger posed to represen-
tative government by initiative campaigns.10
He lamented the power of money in initiative
efforts and concluded that powerful and
wealthy special interests had taken over the
initiative process. 11 Similarly, the liberal jour-
nalist Peter Schrag noted:
In big states like California, it’s impos-
sible to even get to the table without $1
million or more to pay the lawyers,
consultants, media experts, and, most
important, the signature collectors.
Conversely, the political consultants
who run initiative campaigns also
guarantee that with enough money
behind it, almost any measure can at
least qualify for the ballot.12
3
Writers on the left have also lamented the
restraints on taxation brought by initiatives
and suggested that the traditional American
skepticism of government has become “an
unhealthy distrust.”13
This brief history provides defenders of lim-
ited government no clear guidance about initia-
tives. James Madison’s concerns about majority
rule and his aspirations for representative
democracy are persuasive, especially consider-
ing the Progressive hope to expand government
through direct democracy. Yet the initiative
movement has done much to revive hopes for
limited government in the states.
This study informs the larger debate by
offering an empirical analysis of the effects of
initiatives in the states. The evidence shows
that initiatives have been more effective than Proposition 13
state legislatures at limiting the growth of reflected the will
state and local government spending. That of a broad major-
experience points to some recommendations
for activists seeking to limit government ity of voters in
through the initiative. California in
1978.
Do TELs Work?
During the tax revolt in the 1970s, a num-
ber of states adopted TELs as a mechanism
to limit the growth of government. TELs
place a limit on how much taxes, revenues, or
expenditures can increase in a state each year.
By 1982 TELs had been enacted in 17 states.
TEL enactment slowed along with the fervor
of the tax revolt in the prosperous 1980s;
only 3 states enacted TELs between 1983 and
1990. However, during the early and mid-
1990s, TELs enjoyed a resurgence of sorts; by
1996, 6 additional states had enacted TELs.
Currently, 26 states operate under some kind
of TEL (Table 1).
Many people have questioned the effective-
ness of TELs. Some argue that loopholes and
accounting tricks will inevitably limit the abil-
ity of these measures to reduce government
spending.14 Others say that TELs’ lack of sanc-
tions for noncompliance raises serious ques-
tions about their enforceability. Several stud-
ies that analyze the impact of TELs on state
Table 1
State Tax and Expenditure Limits, 1976–96
Immediate Expenditures Held to
Refunds of Inflation Rate
State Year Passed Surpluses Plus Population Growth

TELs Enacted through Citizen Initiatives

Michigan 1978 Yes No

California 1979 a No No
Washington 1979 No No
Missouri 1980 Yes No
Massachusetts 1986 No No
Colorado 1992 Yes Yes
Washington 1993 No Yes

TELs Passed by State Legislatures

New Jersey 1976 b No No

Colorado 1977 No No
Louisiana 1979 No No
Oregon 1979 Yes No
Idaho 1980 No No
Montana 1981 No No
Utah 1989 No No
New Jersey 1990 No No
Colorado 1991 No No
North Carolina 1991 No No
Iowa 1992 No No
Mississippi 1992 No No

TELs Enacted through Constitutional Conventions

Hawaii 1978 No No
Tennessee 1978 No No

TELs Enacted through Referenda

Arizona 1978 No No
Delaware 1978 No No
Texas 1978 No No
South Carolina 1980 No No
Alaska 1982 No Yes
Oklahoma 1985 No No
Louisiana 1993 No No
Connecticut 1992 No No
Rhode Island 1992 No No
Florida 1994 No No

Source: Mandy Rafool, “State Tax and Expenditure Limits,” National Conference of State Legislatures Legislative
Finance Paper no. 104, 1996, p. 6.

Notes: The TEL that New Jersey passed in 1976 and the TEL that Colorado passed in 1977 are included even though
neither appears in the 1996 NCSL survey. In addition, the TEL that Nevada passed in 1979 is excluded because it
applies only to the governor’s budget proposal, not actual spending and taxing. Referenda differ from initiatives.
Citizens and interest groups can place initiatives directly on the ballot for voter approval. Referenda must be approved
by the state legislature before they appear on the ballot.
a New Jersey’s first TEL expired in 1983.
b In 1979 California passed a TEL that limited per capita appropriations of state tax revenues to the inflation rate.
However, this limit did not pertain to total state expenditures. Additionally, this limit was increased in 1988.

4
budgetary outcomes argue that these mea-
sures do not have a statistically significant
influence on state expenditures.
Those studies, however, have three main
shortcomings. First, the studies often examine
a small number of states15 for a short time,
sometimes as little as a year.16 It is difficult to
draw conclusions from a limited sample; in
any case, the effects of TELs may become
apparent only over an extended time. Second,
the existing studies of TELs often do not take
into account all the factors affecting a state’s
budget.17 Third, previous work does not take
into account the differences among TELs.18
Some TELs limit taxes, some expenditures;
some have strict limits, others looser ceilings
including waiver provisions; some are statuto-
ry, others constitutional. Analyzing the vari-
ous features of TELs might well inform future
efforts to limit spending and taxing.
Table 1 shows that more than two-thirds
of current TELs were enacted either by initia-
tive or by a state legislature. A comparison of
TELs enacted by initiative with those enacted
by state legislatures might prove instructive.
Public choice analysis would suggest that leg-
islators lack the incentive to constrain their
own behavior. They might pass some kind of
fiscal discipline measure to claim credit in
the short term. However, legislators would
Table 2
Limits TELs Impose on Spending and Revenue
Method of Enactment
Citizen Initiative
Limit (7 TELs)
Inflation rate
plus population
growth 2 (29%)
Income growth 5 (71%)
Other 0 (0%)
Source: Author’s calculations based on Mandy Rafool, “State Tax and Expenditure Limits,” National
Conference of State Legislatures Legislative Finance Paper no. 104, 1996, pp. 28–33.
5
make sure a TEL contained loopholes so that TELs passed by
they would be able to avoid cutting popular initiative do
programs in the long term.
Conversely, fiscal discipline measures passed restrain spending
by citizen initiative should be more effective in and taxing and
limiting state spending. Such initiatives would
likely be drafted by groups interested in reduc-
are more effective
ing spending and would be less likely to contain than those passed
loopholes that would benefit elected officials. by legislatures.
TELs enacted by initiative should serve as an
external constraint on legislatures. Indeed,
TELs passed by initiative do restrain spending
and taxing and are more effective than those
passed by legislatures.
Why the Initiative
Is Effective
An examination of the characteristics of
various TELs suggests why TELs passed by
citizen initiative might be more effective than
those passed by legislatures. This can be seen
in four ways. First, TELs differ in the sorts of
limits they set on revenues and expenditures
(Table 2). A vast majority of the TELs that
have been passed in this country hold
increases in expenditures or revenues to
growth in personal income. However, a small
number of states have adopted TELs that
Legislative Vote
(12 TELs)
0 (0%)
5 (42%)
7 (58%)
 have more stringent limitations; they hold
increases in spending or revenue to the infla-
tion rate plus population growth. Table 1
clearly indicates that TELs passed by initia-
tive are more likely to contain this more
stringent limitation. Twenty-nine percent of
the TELs passed by initiative hold increases
in spending or revenues to the inflation rate
plus population growth, while none of the
TELs enacted by legislatures do.
The second important difference between
TELs passed by legislatures and those passed
by citizen initiative concerns transferring
government responsibilities. One frequent
method of circumventing the limits estab-
lished by a TEL is to devolve various func-
tions of government to the localities.
TELs passed by Although this practice may keep the state
initiative are more budget within the limits of the TEL, it will
stringent than result in budget increases at the local level
and will fail to effectively limit overall taxing
those enacted by or spending. However, some TELs include
legislatures. provisions that would remedy this problem.
They include a provision that would man-
date automatic reductions in the limit when-
ever a state devolves a function of govern-
ment to the localities (Table 3). Elected offi-
cials would have no incentive to circumvent
TELs though devolution in states where such
provisions were in place. As can be seen from
Table 3, TELs passed by initiative are more
likely to include provisions that mandate
automatic changes in the limits when
responsibility for government programs is
transferred.
Another structural difference between
TELs that are passed by legislatures and
those that are passed by initiative involves
whether the TELs are constitutional or statu-
tory (Table 4). Constitutional TELs should
be more effective than statutory TELs
because they are more difficult to change.
Statutory TELs leave open the possibility
that the legislature will change the definition
of the item limited, often by excluding cer-
tain areas of spending or revenue. In other
cases, the legislators can simply increase the
limit. By definition, all TELs passed by legis-
latures are statutory. However, 56 percent of
TELs passed by citizen initiative are constitu-
tional. This provides further evidence that
TELs passed by initiative are more stringent
than those enacted by legislatures.
The final structural difference between
TELs that are enacted by legislatures and
TELs that are passed through citizen initia-
tives is their provisions for handling surplus-
es. Most TELs mandate that surpluses go
into a reserve fund or call for taxpayer rebates
only if surpluses persist for a number of
years. However, a limited number of TELs
require that all surplus revenues, above a crit-

Table 3
TELs and Their Provisions for Changing Their Limits: Does Limit Automatically
Change When Responsibility for Government Programs Is Transferred?

Method of Enactment

Citizen Initiative Legislative Vote

Change (7 TELS) (12 TELs)

Yes 5 (71%) 4 (33%)

No 2 (29%) 8 (67%)

Source: Author’s calculations based on Mandy Rafool, “State Tax and Expenditure Limits,” National Conference
of State Legislatures Legislative Finance Paper no. 104, 1996, pp. 28–33.

6
Table 4
Legal Status of TELs

Method of Enactment

Citizen Initiative Legislative Vote

Legal Status (7 TELs) (12 TELs)

Constitutional 4 (56%) 0 (0%)

Statutory 3 (44%) 12 (100%)

Source: Author’s calculations from Mandy Rafool, “State Tax and Expenditure Limits,” National
Conference of State Legislatures Legislative Finance Paper no. 104, 1996, pp. 28–33.

ical threshold, be immediately returned to immediate taxpayer refunds (Table 5). TELs passed by
the taxpayers in the form of tax credits. TELs From the analysis above, it appears that citizen initiative
requiring such refunds make it difficult for TELs that are passed by citizen initiative are are more likely to
state governments to generate revenues that likely to possess a variety of structural fea-
exceed the limit and give taxpayers and tures that render them more effective than have provisions
watchdog groups a greater incentive to see TELs passed by legislatures. The next step that mandate
that TELs are enforced. Once again, we see will be to empirically test whether initiative
that TELs passed by citizen initiative are TELs are more effective by analyzing the rele-
immediate tax-
more likely to have provisions that mandate vant budgetary data. payer refunds.

Table 5
TELs and Their Provisions for Handing Surpluses

Method of Enactment

Provision Citizen Initiative Legislative Vote

for Refund (7 TELs) (12 TELs)

Immediate
taxpayer refunds 3 (44%) 1 (8%)
taxpayer refunds not
immediate 4 (56%) 11 (92%)

Source: Author’s calculations based on Mandy Rafool, “State Tax and Expenditure Limits,” National Conference
of State Legislatures Legislative Finance Paper no. 104, 1996, pp. 28–33.

Note: This provision varies in stringency. Colorado’s TEL mandates that all revenues above the limits be returned
to the taxpayers. Missouri’s and Michigan’s require that revenues exceeding the limit by 1 percent be refunded to
taxpayers. Oregon’s TEL requires that excess revenues above 2 percent of the forecast be returned to the taxpay-
ers. Oregon, whose refund provision is the least restrictive of the four, is the state that passed its TEL by legisla-
tive vote.

7
 Per capita state
and local expendi- Analysis
tures will decrease The empirical test of effectiveness of initia-
by $16.29 every tive TELs involves a regression analysis of a new
data set that includes budgetary data from 49
year after a state of the United States for every fiscal year from
has passed a 1972 through 1996, inclusive.19 Regression
TEL by citizen analysis allows us to examine the effects of var-
ious factors on the central concern of this
initiative. paper, the level of spending at the state and
local level. Regression analysis allows us to sort
out the effects of a single variable by holding
constant the effects of all other variables.
Annual per capita state and local direct
general expenditures in constant 1996 dollars
will be used as the dependent variable in this
regression analysis.20 I examine per capita state
and local direct general expenditures because
doing so best demonstrates how TELs affect
the amount that state and local governments
spend on individual taxpayers.
My regression analysis includes several
variables that might affect state and local
spending. Spending might go up, for exam-
ple, because a state experienced growth in its
population of young people who require
more state services and hence more spend-
ing. I have included a variable measuring
growth in the proportion of people between
the ages of 5 and 17. Similarly, since the elder-
ly use state services at a disproportionately
high rate, an aging population might put
upward pressure on the budget. Therefore, I
have included a variable that measures the
growth in the percentage of people over the
age of 65. It is also possible that increases in
population may place a strain on state ser-
vices and result in budgetary increases.
Hence a variable measuring state population
growth is included as well. Finally, state
spending may also increase because of an
economic slowdown. Hence I include vari-
ables indicating each state’s growth in real
per capita personal income and the annual
change in each state’s unemployment rate.
I also include in the regression analysis four
separate variables that indicate whether a state
has passed a TEL by initiative, by the legislature,
8
by a constitutional convention, or by referen-
dum.21 This will allow me to determine if TELs
passed by citizen initiative are indeed more
effective at limiting the growth of spending
than are TELs enacted by state legislatures.
The results of my analysis can be found in
the Appendix to this paper as Regression 1.
This regression lends support to my hypoth-
esis. The regression explains over 41 percent
of the variation in the dependent variable,
state and local spending. From the regression
results, we see that increases in the percent-
age of state residents aged 5 to 17 lead to
increases in per capita state and local direct
general expenditures. This finding is statisti-
cally significant.22 In addition, as a state’s real
per capita personal income grows and as its
unemployment rate declines, there is evi-
dence that per capita state and local expendi-
tures will decrease. Surprisingly, there is evi-
dence that per capita expenditures decrease
as the percentage of state residents over the
age of 65 goes up. However, this finding fails
to achieve statistical significance.
Of more interest in this study are the effects
of the different kinds of TELs. These results are
summarized in Figure 1. The model predicts
that, if other factors are held constant, per capi-
ta state and local expenditures will decrease by
$16.29 every year after a state has passed a TEL
by citizen initiative. Conversely, the model pre-
dicts that TELs enacted by state legislatures will
actually cause per capita expenditures to
increase by $14.00. This finding lends support to
the hypothesis that TELs enacted by citizen ini-
tiative are more effective at limiting state spend-
ing than are those passed into law by state leg-
islatures.
Still, some TELs may be more effective
than others. Perhaps the answer lies in the
provisions of the TELs themselves. Earlier I
demonstrated that TELs passed by initiative
are more likely to have properties that would
make them more effective; however, not all
TELs passed by citizen initiative have those
properties. If we can isolate TELs that have
desirable properties, perhaps we can deter-
mine if a well-designed TEL can place even
further limits on state spending.
Figure 1
Effectiveness of TELs by Method of Enactment
30
25
20
InitiativeTEL: A TEL that is enacted by a citizen initiative
LegislativeTEL: A TEL that is enacted by a state legislature
15
10
-5
-10
-15
-16.29
-20
InitiativeTEL
Two Properties That
Strengthen TELs
Property 1: Limiting the Growth of
Expenditures and Revenues to the
Inflation Rate Plus Population Growth
An overwhelming majority of the TELs that
have been passed since 1976 limit growth in
state expenditures and revenues to state per-
sonal income growth. However, two states,
Colorado and Washington, have recently enact-
ed TELs that limit growth in state expenditures
to the inflation rate plus population growth.
This is a more stringent limit. Over the years the
rate of growth in personal income has been sig-
nificantly greater than the inflation rate.
Between 1980 and 1990 growth in real person-
al income exceeded the inflation rate plus pop-
ulation growth by more than 38 percentage
points.23 It should also be noted that holding
increases in expenditures to increases in per-
sonal income, as most TELs do, sets a relatively
low limit for a state to maintain. Between 1980
9
24
LegislativeTEL
and 1990 the ratio of state and local direct gen-
eral expenditures to personal income actually
fell in 27 of the 49 states considered in this
analysis.24 As a result, TELs that limit growth in
revenues and expenditures to the inflation rate
plus population growth might be more effec-
tive at limiting state and local spending.
Refund provi-
Property 2: Refunding Surpluses to
Taxpayers Immediately sions create a
Another feature that is worth examining strong incentive
is the provision that mandates immediate for state legisla-
refunds of any surpluses to the taxpayers.
Thus far, four states (Colorado, Michigan, tors to cut taxes
Missouri, and Oregon) have enacted TELs when it appears
that mandate immediate refunds of revenues
that exceed the limit established by the TEL.
that revenues are
Three of those four TELs were passed going to exceed
through citizen initiatives; only one the limit.
(Oregon’s) was enacted by the state legisla-
ture. As indicated earlier in this paper, such a
provision would strengthen any TEL because
it would make it difficult for the state gov-
 TELs that limit ernment to collect or spend excess revenues.
increases in In addition, it would give citizens and watch-
dog groups a greater incentive to see that the
spending and provisions of the TEL were enforced. An
revenue to the examination of the recent budgetary history
of these four states indicates that such
inflation rate refund provisions enhance the effectiveness
plus population of TELs in another way. Namely, they create a
growth have the strong incentive for state legislators to cut
taxes when it appears that revenues are going
most promise to exceed the limit.
for reducing If a state enacts a TEL that mandates
spending. immediate refunds of surplus revenues, state
legislators have the option of allowing rev-
enues to exceed the limit and then subse-
quently refunding the revenue. However, there
are logistical and political problems with
doing this. First, it is nearly impossible to
refund the sales tax. Also, although it is possi-
ble to enact refunds of income or property
taxes, legislators dislike doing so. This is
because high-income citizens would obtain a
high percentage of the refunds, and legislators
do not like to be charged with favoring the
rich over everyone else. As a result, this creates
a powerful incentive for legislators to cut taxes
so that revenues or expenditures will no longer
exceed the limit. Indeed, case studies indicate
that Michigan, Missouri, and Colorado (three
of the four states that mandate taxpayer
refunds) have enacted tax cuts in response to
the prospect of having revenues exceed the
limit mandated by their TELs.25
To further study these provisions, I have
constructed another regression model. Once
again, the dependent variable is the annual
change in per capita state and local direct
general expenditures in 1996 dollars. The
analysis examines the effects of TELs with
provisions for immediate refunds and TELs
that limit growth in expenditures to the
inflation rate. The demographic and eco-
nomic variables included in the first regres-
sion are included in this regression as well.
The regression results can be found in the
Appendix as Regression 2.
These results, which are summarized in
Figure 2, support the idea that certain fea-
tures can greatly enhance the effectiveness of
10
a TEL. From this regression it appears that
TELs that limit increases in spending and
revenue to the inflation rate plus population
growth have the most promise for reducing
spending. If a state passes a TEL that limits
expenditures to the inflation rate plus popu-
lation growth, the regression equation pre-
dicts that every year the TEL will reduce per
capita state and local direct general expendi-
tures by approximately $114.84. The t-statis-
tic indicates that we can be more than 98 per-
cent confident that these TELs have a nega-
tive effect on state and local direct general
expenditures. Likewise, if a state passes a TEL
that does not limit state expenditures to the
inflation rate plus population growth but
includes a refund provision, the regression
equation predicts that the TEL will reduce
per capita direct general expenditures by
$39.80 annually.
Finally, the regression analysis suggests
that other TELs that neither limit expendi-
tures to inflation nor have immediate refund
provisions appear ineffective at reducing
state expenditures. The model predicts that if
a state passes a TEL that has neither of these
two provisions, that state’s per capita direct
general expenditures will actually increase by
$14.59. Overall, this analysis provides strong
evidence that TELs can be effective tools for
limiting the growth of state expenditures,
but only if they are designed properly.
Moreover, since state legislatures generally
lack incentives to constrain their own behav-
ior, TELs passed by citizen initiatives are far
more likely to contain the sorts of provisions
that are going to place effective limits on
state spending.
A Closer Look at
Washington and Colorado
In general, these results seem very promising.
It appears that the two TELs passed by
Washington and Colorado that limit the
growth of expenditures to the inflation rate
have been especially effective at reducing per
capita government expenditures. Unfortunate-
Figure 2
Effectiveness of TELs by Feature

40

20 14.59

-20

-40
-39.80
-60
StrongTEL: A TEL that limits per capita expenditure growth to the inflation rate
RefundTEL: A TEL that is not a StrongTEL but does mandate immediate
-80 refunds during times of surplus
OtherTEL: A TEL that is neither a StrongTEL nor a RefundTEL
-100

-120 -114.84

-140
StrongTEL RefundTEL OtherTEL

ly, however, there are few data to use to analyze
the impact of these TELs. Colorado’s TEL was
passed in 1992 and took effect in FY94, and
Washington’s TEL was passed in 1993 and took
effect in FY96.26 Since 1996 is the most recent
year for which the Census Bureau provides data
on state and local expenditures, I have only one
year’s data to use to analyze the impact of
Washington’s TEL and three years’ data to use
to examine Colorado’s TEL. As a result, in order
to further this analysis I will have to use anecdo-
tal evidence in both cases.
Case Study: Colorado
In the past 25 years Colorado has enacted
three separate TELs. In 1977 Colorado was
one of the first states to adopt a general fund
appropriations limit. The legislation limited
increases in state appropriations to 7 percent
over the previous year’s general fund appro-
priations. Due to expire after FY83, the law
was amended in 1979 and extended indefi-
nitely. However, during the mid to late 1980s
Colorado’s economy suffered a downturn
due to the collapse of the energy and con-
struction industries, and revenues were con-
sistently below the limit mandated by the
TEL. In 1991 the General Assembly of
Colorado adopted another statutory general
fund appropriations limit. This one reduced
the existing limit by one percentage point,
mandating that general fund expenditures
could increase by no more than 6 percent. Colorado’s citi-
However, this legislation included generous
exemptions for spending on education and zens became
federal mandates.27 increasingly frus-
Colorado’s citizens became increasingly frus- trated by what
trated by what they believed to be government
inefficiency and the perceived inequities in the they believed to
state tax system. Many became involved with a be government
grassroots movement to reform state and local
taxes. In 1986, 1988, and 1990 they succeeded in
inefficiency and
placing on the ballot initiatives that would limit the perceived
taxes and spending. Those initiatives lost by nar- inequities in the
rower margins each time. Finally, in 1992, the
Taxpayer Bill of Rights (TABOR), also known as state tax system.
Amendment One, passed and added Article X,
sec. 20, to the state constitution.28

11
Most people agree TABOR has three primary components.
that the dire First, all tax increases have to be approved by
taxpayers. Second, it mandates that the exist-
predictions of ing TELs, passed in 1977 and 1991, cannot be
opponents of weakened without taxpayer approval. Third,
it includes the most stringent TEL of any
TABOR have not state. TABOR limits growth in state spend-
come to pass. ing and tax increases to inflation plus popu-
lation growth.29 It mandates that any revenue
collected over the limit be refunded to the
taxpayers. It requires that the limit be adjust-
ed when responsibility for government pro-
grams is transferred. Finally, the limit is con-
stitutional, not statutory, which makes it dif-
ficult to amend.
This particular ballot initiative generated a
firestorm of controversy. Gov. Roy Romer, a
Democrat, sharply criticized the measure on
numerous occasions. He said that defeating
the measure was the “moral equivalent of
fighting the Nazis at the Battle of the Bulge.”
He warned of an economic Armageddon with
passage of TABOR and said that the Colorado
border would soon have to be posted with
signs reading “Colorado is closed for busi-
ness.”30 Public employee unions and the edu-
cation lobby quickly lined up in opposition to
TABOR. Even the New York Times criticized
TABOR, calling it potentially the most radical
change in any state government that year. 31
Others argued that TABOR was bad policy
because the demands for state services, such as
schools, prisons, and highways, seemed likely
to increase faster than the rate of inflation.
They also contended that Colorado needed to
spend more on those services because, in pre-
vious years, Colorado’s spending increases for
education and highways had been consider-
ably below the national average.32 Despite
those warnings, TABOR passed with more
than 53 percent of the vote in 1992 and took
effect in FY94.
Since 1994 the legislature has had to
rebate substantial amounts of tax revenues
to stay underneath the limit. Colorado enact-
ed taxpayer refunds of $139 million in 1997,
$563 million in 1998, $679 million in 1999,
and $941 million in 2000.33 In addition to its
rebate provisions, TABOR forces both state
12
and local government to obtain voter
approval to raise taxes. Although many
municipalities have sought and won voter
approval to increase taxes, 34 statewide initia-
tives have fared poorly. In every year from
1993 to 1999 a proposal to either increase
taxes or circumvent TABOR was on the
Colorado ballot. Those included a 1993 ini-
tiative to increase the sales tax, a 1997 gas tax
increase, and a 1999 effort to use part of the
surplus for road and school construction.
Each of those statewide initiatives was defeat-
ed.35 However, in 2000 Colorado residents
did approve Amendment 23, which increased
state aid to public education and reduced the
TABOR surplus for both 2000 and 2001.36
Most people agree that the dire predictions
of opponents of TABOR have not come to
pass. However, many argue that when the
economy slows down, it will become more
difficult for the state legislature to stay with-
in the limit.37
Case Study: Washington State
During the past 20 years Washington
State has passed two TELs by citizen initia-
tive. The first one, Initiative 62, was passed in
1979 and limited increases in state revenues
to the rate of growth in personal income.
However, the state suffered a recession short-
ly after passage of the initiative, and it never
became a serious constraint since the limit
was higher than what the state could spend.
In fact, in 1993 the legislature was able to
pass a $1 billion tax increase to balance the
FY94–95 budget and remain within the limit.
However, that tax increase provoked a back-
lash and provided the impetus for putting
another TEL, Initiative 601, on the ballot in
1993.38
Initiative 601 imposed a limit that was more
strict than the limit set by Initiative 62. Initiative
601 limited increases in state expenditures to
the inflation rate. In addition, it stopped the
legislature from circumventing the limit by
devolving functions of government to the local-
ities. It explicitly prohibited the legislature to
impose on local governments any responsibili-
ty for new programs unless the legislature fully
reimbursed the local governments for the cost
of the programs. Initiative 601 passed by 1 per-
cent of the vote.39
In 1994 the Washington legislature
passed a supplemental budget to ensure that
it was in compliance with the TEL that was
scheduled to take effect in FY96. The legisla-
ture instituted some targeted budget cuts,
mostly in administration, social services, and
prisons, to save more than $120 million in
the new biennium. The legislature increased
spending for some items, such as highways
and school construction, on the grounds
that those were one-time-only expenses and
would be off budget in FY96. As a result, the
budget base was not swollen from previous
spending levels and would be easier to sus-
tain in the new biennium. Finally, some agen-
cies were directed to begin planning for cuts.
For instance, public colleges were directed to
trim expenses by $39 million to help pay for
faculty and staff pay raises. 40
Because of those spending reductions, the
budget was under the TEL’s limit in FY96
and FY97. In subsequent years the state legis-
lature took steps to reduce taxes when it
appeared that the government was collecting
high levels of revenue. In FY98 and FY99 the
legislature instituted modest targeted tax
cuts of $38.5 million and $19.7 million,
respectively.41 Since spending was being
restrained, voters in the state of Washington
desired more substantial tax relief. In 1998
Washington voters passed Initiative 695,
which reduced the motor vehicle excise tax by
$30 and saved taxpayers $256 million. In
1999 Washington residents voted to repeal
the motor vehicle excise tax. That reduced
the tax burden on Washington residents by
an additional $1.1 billion.42
However, Washington’s Initiative 601 was
weaker than Colorado’s Taxpayer Bill of
Rights in one important respect. Initiative
601 was a statutory measure whereas TABOR
13
is a constitutional amendment. This makes TELs that impose
Initiative 601 easier to amend, and possibly more stringent
weaken. Indeed, that is precisely what hap-
pened in the spring of 2000 when the limits on expen-
Washington legislature wanted to pass a ditures are also
budget that would have exceeded the limit
mandated by the TEL. The legislature suc-
effective at
ceeded in obtaining the necessary superma- restraining
jority to suspend the TEL, and the governor government
signed the budget into law.43 The long-term
effects on the budgetary practices of the state growth.
of Washington remain to be seen.
Conclusion
I have used regression analysis to examine
the impact of various TELs on state and local
expenditures. The approach is comprehen-
sive, as budgetary data from 49 states for
more than 25 years are used. Most impor-
tant, special attention is paid to the manner
in which the TELs were enacted. This pro-
vides a number of interesting insights. The
analysis indicates that TELs passed by citizen
initiative procedures are more effective in
limiting state spending and revenues than
are TELs passed by legislatures.
Certain features make some TELs more
effective than others. There is solid evidence
that TELs that require immediate taxpayer
refunds of surpluses are effective in reducing
expenditures. There is even stronger evidence
that TELs that impose more stringent limits
on expenditures are also effective at restrain-
ing government growth. This is demonstrat-
ed by statistical analysis and case studies of
budgetary outcomes in Colorado and
Washington State. The data indicate that
activists wishing to restrain government
growth should focus on passing TELs that
include immediate taxpayer refunds of sur-
pluses and hold spending increases to the
inflation rate plus population growth.
 Appendix: Regression Results
Regression 1
Technique: WLS, panel corrected standard errors fixed effects model with state and year
indicator variables.
Dependent variable: Annual change in per capita state and local direct general expendi-
tures in 1996 dollars for 49 states from 1972 through 1996.

Multiple R .642
R square .413
Standard error 111.092

coef s.e. t-stat sig t

UnemploymentCH 2.88 4.90 0.76 .4474

Age5to17CH 59.93 27.35 2.19 .0287*
Age65CH -1.67 13.93 0.12 .9045
PopulationGrowth 43.69 5.19 8.91 .0000*
IncomeGrowth -2.67 1.41 -1.89 .0590
InitiativeTEL -16.29 20.05 -0.81 .4181
LegislativeTEL 14.00 16.59 0.84 .4011
ConventionTEL 61.67 38.97 1.58 .1144
ReferendumTEL -1.95 18.59 -0.10 .9204
Constant 165.07 29.23 5.65 .0000*

*Statistically significant.
Regression 2
Technique: WLS, panel corrected standard errors fixed effects model with state and year
indicator variables.
Dependent variable: Annual change in per capita state and local direct general expendi-
tures in 1996 dollars for 49 states from 1972 through 1996.

Multiple R .646
R square .417
Standard error 108.922

coef s.e. t-stat sig t

UnemploymentCH 2.59 3.78 0.68 .4966*

Age5to17CH 63.22 27.48 2.30 .0216*
Age65CH -2.31 13.99 -0.16 .8729
PopulationGrowth 43.29 4.95 8.75 .0000*
IncomeGrowth -2.61 1.41 -1.85 .0645
StrongTEL -114.84 53.57 -2.14 .0326*
RefundTEL -39.80 24.04 -1.66 .0972
OtherTEL 14.59 12.61 1.16 .2463
Constant 172.82 24.37 7.09 .0000*

*Statistically significant.

14
Explanation of Dependent Variables
UnemploymentCH indicates the annual
percentage point change in the state’s unem-
ployment rate.
Age5to17CH indicates the annual percent-
age point change in the number of state resi-
dents who are between the ages of 5 and 17.
Age65CH indicates the annual percentage
point change in the number of state residents
who are over the age of 65.
PopulationGrowth indicates the growth
in population in each state.
IncomeGrowth indicates the annual
change in per capita personal income in each
state, adjusted for inflation.
StrongTEL is an indicator variable that is
scored a 1 if a state has passed a TEL that lim-
its growth in per capita expenditures to the
inflation rate, zero otherwise.
RefundTEL is an indicator variable that is
scored a 1 if a state has passed a TEL that is
not a StrongTEL but does mandate immedi-
ate taxpayer refunds during times of surplus,
zero otherwise.
OtherTEL is an indicator variables that is scored
a 1 if a state has passed a TEL that is neither a
StrongTEL nor a RefundTEL, zero otherwise.
Notes to Appendix
The reported statistical significance for
the coefficients is two tailed.
Data are weighted to correct for het-
eroskedasticity.
Both a White test and a Breusch Pagan
Godfrey test indicate that there is no statisti-
cally significant evidence of heteroskedastici-
ty in the residuals. An Asymptotic test indi-
cates that there is no statistically significant
evidence of serial correlation in the residuals.
The technique that Nathaniel Beck and
Jonathan Katz describe in “What to Do (And
Not to Do) with Time-Series Cross-Sectional
Data,” American Political Science Review 89, no.
3 (September 1995): 634–47, is used to do the
panel corrections.
Notes
1. David S. Broder, Democracy Derailed: Initiative
15
Campaigns and the Power of Money (New York:
Harcourt, 2000), p. 7.
2. Peter Schrag, “Rule by Referendum,” American
Prospect, July 17, 2000, p. 38.
3. Properly defined, Proposition 13 is not a state-
wide tax and expenditure limitation but a property
tax limitation. However, its passage did mark the
beginning of the tax revolt, which is why it is men-
tioned here.
4. James Madison, Federalist no. 10 (1787), in The
Founders’ Constitution, ed. Philip B. Kurland and
Ralph Lerner. (Chicago: University of Chicago Press,
1978), vol. 1, pp. 128–30. Alexander Hamilton at the
Constitutional Convention was even less favorable
toward direct democracy: “It has been observed, by
an honorable gentleman, that a pure democracy, if it
were practicable, would be the most perfect gov-
ernment. Experience has proved that no position in
politics is more false than this. The ancient democ-
racies, in which the people themselves deliberated,
never possessed one feature of good government.
Their very character was tyranny; their figure, defor-
mity. When they assembled, the field of debate pre-
sented an ungovernable mob, not only incapable of
deliberation, but prepared for every enormity.”
Debates on the Adoption of the Federal Constitution, ed.
Jonathan Elliott (Washington: privately printed,
1836), vol. 2, p. 253.
5. David Magleby, Direct Legislation: Voting on Ballot
Propositions in the United States (Baltimore: Johns
Hopkins University Press, 1984), p. 22. See also
John M. Allswang, The Initiative and Referendum in
California 1898–1998 (Stanford: Stanford Univer-
sity Press, 2000).
6. Quoted in Magleby, p. 23.
7. Robert Higgs, Crisis and Leviathan: Critical
Episodes in the Growth of American Government (New
York: Oxford University Press, 1987), passim.
8. Allswang, pp. 105–7.
9. Ibid., pp. 108–9.
10. Broder, p. 16.
11. Ibid., chap. 4.
12. Schrag, “Rule by Referendum,” p. 38.
13. Broder, p. 20. See also Peter Schrag, Paradise
Lost: California’s Experience, America’s Future (New
York: New Press, 1998).
14. James Bennett and Thomas DiLorenzo, “Off-
Budget Activities of Local Governments: The Bane of
the Tax Revolt,” Public Choice 39, no. 3 (1982): 333–34.
15. James Cox and David Lowery, “The Impact of
Tax Revolt Era State Fiscal Caps,” Social Science
Quarterly 3 (1990): 492–509; and David Lowery
and Tyson King-Meadows, “The Impact of Tax
Revolt Era State Fiscal Caps: A Research Update,”
Public Budgeting and Finance 16, no. 1 (1996):
102–12.
16. Burton Abrams and William Dougan, “The
Effects of Constitutional Restraint on Govern-
ment Spending,” Public Choice 49, no. 2 (1986):
101–16; and Dale Bails and Margaret Tieslau,
“The Impact of Fiscal Constitution on State and
Local Expenditures,” Cato Journal 20, no. 2 (2000):
255–77.
17. Marcia Howard, “State Tax and Expenditure
Limitations: There Is No Story,” Public Budgeting
and Finance 9, no. 2 (1989): 83–90; Dean Stansel,
“Taming Leviathan: Are State Tax and Spending
Limits the Answer?” Cato Institute Policy
Analysis no. 213, July 25, 1994; Dale Bails, “The
Effectiveness of Tax and Expenditure Limits: A
Re-evaluation,” American Journal of Economics and
Sociology 49, no. 2 (1990): 223; and Daphne
Kenyon and Karen Benker, “Fiscal Discipline:
Lessons from the State Experience,” National Tax
Journal 37, no. 3 (1984): 433–46.
18. Phillip Joyce and Dan Mullins, “The
Changing Fiscal Structure of the State and Local
Sector: The Impact of Tax and Expenditure
Limitations,” Public Administration Review 51, no. 3
(1991): 240–52; Phillip Joyce and Dan Mullins,
“Tax and Expenditure Limitations and State and
Local Fiscal Structure: An Empirical Assessment,”
Public Budgeting and Finance 16, no. 1 (1996):
75–101; and Ronald Shadbegian, “Do Tax and
Expenditure Limitations Affect the Size and
Growth of Government?” Contemporary Economic
Policy 14, no. 2 (1996): 22–35.
19. Alaska receives a high percentage of its rev-
enue from severance taxes on oil produced and
minerals mined in the state. Since shifts in the
prices of those commodities cause a great deal of
fluctuation in Alaska’s budgetary outcomes,
Alaska is omitted from this analysis.
20. It should also be noted that in this analysis
combined state and local data will be used instead
of state data because the combined data tend to
be more stable. In addition, state and local data
provide a more accurate indication of the level of
expenditures in each state. As was mentioned ear-
lier, some states are able to create phantom reduc-
tions in taxes and spending without reducing the
overall tax burden by simply devolving functions
of state government to the localities. That could
yield some potentially misleading results if only
state data were used. However, this problem is
16
neatly avoided by using data that combine state
and local expenditures.
21. However, in this paper no hypotheses are going
to be posed about TELs passed by constitutional
conventions for a couple reasons. First, only two
states in this analysis, Tennessee and Hawaii, have
passed a TEL though constitutional convention,
and, as a result, there is a paucity of data. Second, I
lack the details about these particular conventions
needed to make an informed hypothesis about the
strength of the TELs. If most of the delegates to
constitutional conventions are legislators, then I
would expect the TEL to be weak. However, if most
of the delegates are not elected officials, there exists
a possibility that a more stringent TEL may be
enacted. However, I do not possess the necessary
details about these particular conventions to make
an informed judgment.
Likewise, no hypotheses are going to be posed
about TELs that are enacted through referendum
procedures. Since a referendum must be
approved by the legislature before being placed
on the ballot, it is difficult to predict the effec-
tiveness of this sort of TEL. It is reasonable to
assume that a legislature would be unlikely to
approve a ballot proposal that would impose
severe restrictions on its behavior. However, since
there is a possibility that the referendum will fail
to get the necessary majority, legislators might be
more willing to place a restrictive TEL on the bal-
lot than to enact a restrictive TEL directly. As a
result, I would anticipate that TELs passed by ref-
erendum would be less effective than those
passed by initiative but more effective than those
passed by legislatures. However, a formal hypoth-
esis about this kind of TEL will not be put forth.
22. Saying that a variable is statistically significant
means that we can be at least 95 percent certain that
the given variable actually has a nonzero effect on state
and local expenditures. In other words, the probability
that chance variation within the data caused an effect
as great or greater is 5 percent or less.
23. Bureau of the Census, Statistical Abstract of the
United States: 2000 (Washington: Government
Printing Office, 2000), p. 485; Bureau of the
Census, Statistical Abstract of the United States: 1992
(Washington: Government Printing Office,
1992), p. 438; and Bureau of the Census, Statistical
Abstract of the United States: 1991 (Washington:
Government Printing Office, 1991), p. 22.
24. Bureau of the Census, Governmental Finances
(Washington: Government Printing Office, 1980
and 1990).
25. Mandy Rafool, “State Tax and Expenditure
Limitations,” National Conference of State
Legislatures Legislative Finance Paper no. 104,
1996, p. 14; David Webber, “The Missouri
Experience with State Revenue Limits—The
Hancock Amendment,” Paper presented at Policy
Collaborative on Tax and Spending Limits in the
States, Colorado University, Denver, July 2001, p.
13; and Franklin James, “Tax and Spending
Limits on Colorado,” Paper presented at Policy
Collaborative on Tax and Spending Limits in the
States, Colorado University, Denver, July 2001, p.
10.
www.revenue.state.co.us/refund2000.html;
www.revenue.state.co.us/refund99.html;
www.revenue.state.co.us/refund98.html; and
www.revenue.state.co.us/refund.html.
34. John Sanko, “’92 Election Was Fiscal Face Lift,”
Rocky Mountain News, August 3, 1999, p. 14A.
35. “TABOR Legislative Handbook,” Indepen-
dence Institute Paper, 1999, pp. 1–2.

26. Rafool, “State Tax and Expenditure Limits,” p. 18. 36. James, p. 11.

27. Ibid., p. 13. 37. Rafool, “State Tax and Expenditure Limits,” p. 14.

28. Ibid. 38. Ibid., p. 17.

29. Rafool, “State Tax and Expenditure Limita-
tions,” p. 28.
30. Stephen Moore and Dean Stansel, “The Great
Tax Revolt of 1994,” Reason, October 1994, p. 20.
31. Dirk Johnson, “Taxpayer Revolt in Colorado
Raises Alarm about Lost Services,” New York
Times, November 15, 1992, p. 18.
39. Ibid., p. 18.
40. Ibid.
41. Judy Zelio, State Tax Actions 1997 (Denver:
National Conference of State Legislatures, 1998),
p. 15; and Judy Zelio, Scott Mackey, and Mandy
Rafool, State Tax Actions 1998 (Denver: National
Conference of State Legislatures, 1999), p. 11.

32. Nicholas Johnson, “The Taxpayer Bill of Rights
and Referendum B: Assessing the Availability of
Public Services and Investment,” Center for Budget
and Policy Priorities, Washington, 1998, p. 3–7.
33. Colorado Department of Revenue Web sites:
42. Mandy Rafool, State Tax Actions 1999 (Denver, Colo:
National Conference of State Legislatures, 2000), p. 39.
43. Bob Williams, “House Budget Shreds Spending
Limit (I-601),” Evergreen Freedom Foundation,
Olympia, Washington, Policy Highlighter, 2000.

Published by the Cato Institute, Policy Analysis is a regular series evaluating government policies and offer-
ing proposals for reform. Nothing in Policy Analysis should be construed as necessarily reflecting the views
of the Cato Institute or as an attempt to aid or hinder the passage of any bill before congress. Contact the
Cato Institute for reprint permission. Additional copies of Policy Analysis are $6.00 each ($3.00 each for five
or more). To order, or for a complete listing of available studies, write the Cato Institute, 1000 Massachusetts
Ave., N.W., Washington, D.C. 20001, call toll free 1-800-767-1241 (noon - 9 p.m. eastern time), fax (202) 842-
3490, or visit our website at www.cato.org.

17